随机森林算法在中药指纹图谱中的应用:以不同品牌夏桑菊颗粒指纹图谱分析为例

发布时间：2019-01-17 12:57

【摘要】：该研究旨在建立随机森林算法鉴别和分类不同品牌夏桑菊颗粒,为多指标的复杂指纹图谱的鉴别提供有效的参考。采用高效液相法采集83批不同品牌的夏桑菊颗粒指纹图谱,比较主成分分析、偏最小二乘法-判别分析、随机森林等方法在处理不同分类样品复杂数据时的不同。结果表明本研究成功建立了83批不同品牌夏桑菊颗粒的指纹图谱;经过不同模式识别方法比较可得,主成分分析分析只能解释56.52%的方差贡献率,同时不能完全将样品分类;偏最小二乘法-判别分析优于主成分分析的结果,能达到一定的分离,解释总体方差贡献率63.43%;而随机森林法能够很好的将样品分为3类,且3类样本的10折交互验证准确率达到96.5%。因此,随机森林算法联合HPLC指纹图谱能够有效构建中药质量控制和分析体系。
[Abstract]:The purpose of this study is to establish a random forest algorithm for identifying and classifying different brands of Xiasangju granules, and to provide an effective reference for the identification of complex fingerprints with multiple indexes. High performance liquid phase (HPLC) method was used to collect 83 samples from different brands of Xiasangju granule fingerprint, and to compare the different methods of principal component analysis, partial least squares discriminant analysis and random forest in dealing with complex data of different classification samples. The results showed that the fingerprint of 83 different brands of Xiasangju granules was successfully established in this study. Compared with different pattern recognition methods, principal component analysis can only explain 56.52% variance contribution rate, and can not completely classify samples. Partial least squares discriminant analysis is superior to principal component analysis and can achieve a certain separation and explain the total variance contribution rate of 63.43. The random forest method can divide the samples into three categories, and the accuracy of 10% cross-validation of the three types of samples is 96.5. Therefore, the random forest algorithm combined with HPLC fingerprint can effectively construct the quality control and analysis system of traditional Chinese medicine.
【作者单位】：湖南中医药大学药学院湘产大宗药材品质评价湖南省重点实验室中药有毒物质防控技术湖南省工程实验室湖湘中药资源保护与利用协同创新中心;
【基金】：国家“重大新药创制”科技重大专项(2013ZX09201019) 教育部高等学校博士学科点专项科研基金项目(20124323120004) 湖南省自然科学基金项目(13JJ4089) 湖湘青年科技创新创业平台项目(2013) 湖南省十二五重点学科药学学科项目湖南中医药大学人才引进项目(2014RS4009)
【分类号】：R286.0

【相似文献】